基于负熵和智能优化算法的盲源分离方法

Blind Source Separation Method Based on Negative Entropy and Intelligent Optimization Algorithm

针对混合蛙跳算法(SFLA)更新策略会陷入局部最优、降低收敛速度的问题,提出一种自适应阈值更新策略。根据盲源分离中常用峭度和负熵作为非高斯性的度量,但峭度对野值敏感,影响算法性能,研究一种基于负熵准则的采用粒子群优化(PSO)算法和混合蛙跳算法的盲源分离方法。仿真结果表明,基于负熵的盲分离算法性能优于基于峭度的盲分离算法,基于SFLA的盲分离算法性能优于基于PSO的盲分离算法。

Aiming at the problem of slowing the convergence and facing local optimization of Shuffled Frog Leaping Algorithm（SFLA） update strategy, this paper proposes a update strategy of adaptive threshold selection is raised. Kurtosis and negative entropy are used as a measure of non-Gaussian in Blind Source Separation（BSS）, but kurtosis is sensitive to outliers affecting performance of BSS, it researches a criteria of negative entropy based on Particle Swarm Optimization（PSO） algorithm and SFLA. Simulation results show that the proposed BSS of negative entropy has significant performance improvement over BSS of Kurtosis and BSS based on SFLA has better performance over BSS based on PSO.